Integrated circuits are made in a bulk parallel process by patterning and processing semiconductor wafers. Each wafer contains many identical copies of the same integrated circuit referred to as a “die.” It may be preferable to test the semiconductor wafers before the die is cut into individual integrated circuits and packaged for sale. If defects are detected the defective die can be culled before wasting resources packaging a defective part. The individual die can also be tested after they have been cut into individual integrated circuits, either before or after being packaged (here “packaged” refers to the process of being connected to an electrical interconnect package that protects the die and allows for assembly onto a wiring board).
To test a wafer or an individual die—commonly called the device under test or DUT—a probe card is commonly used which comes into contact with the surface of the DUT. The probe card generally contains three unique characteristics: (1) an XY array of individual probes that move in the Z direction to allow contact with the die pad; (2) an electrical interface to connect the card to a circuit test apparatus; and (3) a rigid reference plane defined in such a way that the probe card can be accurately mounted in the proper location. When the probe card is brought in contact with the die pad, the Z-direction movement allows for a solid contact with the probe tip. The probe card ultimately provides an electrical interface that allows a circuit test apparatus to be temporarily connected to the DUT. This method of die testing is extremely efficient because many die can be tested at the same time. To drive this efficiency even higher, probe card manufactures are making larger probe cards with an ever-increasing numbers of probes.
Currently two major types of probe designs are used to test a semiconductor die—cantilever and torsional FIGS. 6A and 6B illustrate a conventional cantilever probe. The probe (605) comprises a probe tip (610), a bending element (615), and a probe base (620), which is mounted to a substrate (625). This entire structure is referred to herein as the probe card. The DUT is generally moved relative to the fixed probe card in the Z-direction (depicted by arrow 630) causing the bending element (615) to bend allowing the probe tip (610) to come into contact with the die pad that is under test. FIG. 6B illustrates how the probe bending element (635) bends while being brought into contact with the die. As an individual probe travels to make contact with the DUT contact pad (this event is called a touchdown), the probe tip scrubs the contact pad, which perfects an electrical contact with the die such that testing can commence. The die contact pads, which are typically aluminum (but may be Gold, Solder, and increasingly Copper), are often coated with a thin layer of aluminum oxide, or other undesired passivating coating, and the probe tip must cut through the coating to perfect the electrical connection. Once testing is complete, the probe (605) is moved away from the die pad and the probe springs back to its original position.
The cantilever design, however, has several shortcomings. Typical cantilever probes are designed with long bending elements, which during touchdown, the probe tip presents a smaller tip contact angle (as compared to the same sized tip on a typical torsion probe design). This results in a larger tip contact area and a larger probe force is thus needed to pierce the aluminum oxide layer. When you multiply this force by the hundreds or thousands of probes on a probe card, the probe card must be engineered to accommodate significant forces, which usually means reinforcing the probe card components, which in turn increases probe card costs and test system complexity.
Another shortcoming is the inefficient distribution of stresses. During touchdown, a cantilever probe bends, which creates stresses on the probe that appear concentrated at the top and bottom surfaces of the bending element near the probe base end of the probe. FIG. 7A illustrates a length-wise cross-sectional view of the stresses experienced by the bending element of a cantilever probe, while FIG. 7B illustrates the width-wise cross-sectional views (Sections A-A and B-B) of the stresses at each end of the element. The left side of the figure, near Section A-A, (indicated by part 705) is the part of the bending element that is near the probe base, with the right side, near Section B-B, (part 710) near the probe tip. The area of the bending element that experiences stresses which are greater than 50% of the maximum stress is shown hatched (815). The corresponding volume of the bending bar that experiences greater than 50% of maximum stress is about 25% of the total cantilever bar volume, and that volume is localized near the probe base (705). The opposite side of the bending bar (710) experiences very low stress. It is clear from FIGS. 7A and 7B that the stress distribution is inefficient because only small portions of the bending element absorb the stress. And it is in these small portions where the probe is more likely to fail forcing manufacturers to strengthen the bending element (often by widening at the probe foot) to reduce stress and prevent failures.
Shown in FIG. 8 is a cantilever probe (805) with a constant width in the bending element (810). As is shown in the dark area (815), the cantilever probe (805) experiences the most amount of stress at the probe base. Cantilever probe (820) in FIG. 8 has a bending element (825) that is wider at the probe base and tapers to the probe tip. As shown by the dark area (830), the stress experienced by the bending element (825) is much less than in the non-widened design. This design, however, is not without costs. A wider bending element near the probe base adversely affects the packing density of the probe card as shown in FIG. 9. The inefficient packing density renders it difficult to scale probe layout and design to test finer pitch DUTs (where the pads are closer together).
The second type of probe is based on a torsional design which was developed to overcome some of these drawbacks. For example, U.S. Pat. No. 6,426,638 describes a torsion spring design. FIG. 6C illustrates a torsional probe design. As the probe tip (81) comes in contact with the DUT contact pad, it moves flexibly in response to force applied vertically to the tip (81). Vertical movement of the tip (81) depresses the arm (82) and torsionally flexes the torsion element (83) in the direction indicated by arrow (90). The torsion element (83) serves as a torsional spring, thereby impressing a restoring force on the tip (81).
Torsional designs have some advantages over cantilever designs. Typical torsional probes are designed with short arms, which during touchdown, the probe tip presents a larger tip contact angle (as compared to the same sized tip on a typical cantilever probe design). This results in a smaller tip contact area and a smaller probe force is thus needed to pierce the aluminum oxide layer, which in turn, reduces the overall force exerted by the probe card. Reduced overall force is advantageous because the probe card does not need to be reinforced as much as a comparable cantilever design, thereby reducing manufacturing costs or allowing more probes to be built on a card.
Finally, the torsional design more efficiently distributes stress across the entire volume of the torsion element. FIG. 10A illustrates a length-wise cross-sectional view of the stresses experienced by the torsion element, while FIG. 10B illustrates the width-wise cross-sectional views (Sections C-C and D-D) of the stresses at each end of the element. The area of the torsion element that experiences stresses that are greater than 50% of the maximum stress is shown hatched area (1005), with the center of the torsion element (1010) experiencing the least amount of stress. The corresponding volume of the torsion element that experiences greater than 50% of maximum stress is about 60% of the total torsion element volume. Unlike a cantilever design, this stress is experienced throughout the entire length of the torsion element, and it is not localized at the probe base. It is therefore more efficient to make the width of the torsion bar uniform, thereby also improving the packing density.
Unfortunately, the torsional probe too has drawbacks. First, for a typical torsional design with a shorter geometry of the arm, the scrub length is generally longer which can limit the size of contact pads for the DUTs. Second, again because of the typically shorter geometry of the arm, a large torsional force is required for larger z-displacement of the probe tip during probing. This large torsion may result in small z-deformation in the torsion element caused by material fatigue, which can translate into a z-shift. Third, the torsional probe also limits packing density in some DUT layout configurations. To illustrate, FIG. 11A presents a torsional probe configuration (1105) with the probe base (1110) and probe tips (1115). In this configuration, a probe card can effectively test a DUT that has a contact pad configuration that matches the probe tip configuration. In FIG. 11B, a new probe configuration (1120) is presented which has a different probe tip configuration (1125). However, this configuration would not work because in at least two areas (1130 and 1135) the probes will obstruct each other and prevent the probe card from operating efficiently.
What is needed, therefore, is a probe that exploits the advantages of both the torsional and cantilever probe designs while reducing their associated shortcomings.